Timeline reactions: market selloff after Nvidia beat, Michael Burr's AI circularity skepticism, and Bitcoin down 10%
Nov 20, 2025
Key Points
- Nvidia posts record earnings and guides higher, yet NASDAQ falls 2.1% and Bitcoin drops 5%, signaling investor skepticism about AI's path to profitability.
- Michael Burr argues AI revenue is circular—venture and corporate funding flowing back to the same ecosystem rather than genuine end-user demand—challenging the industry's core growth narrative.
- AI capability gains track a steady exponential curve with no acceleration yet, leaving the critical question of whether current infrastructure spending will justify itself unresolved.
Summary
Nvidia beats but the market sells off anyway.
Nvidia posted one of the biggest sequential growth quarters in semiconductor history and guided higher. The NASDAQ fell 2.1%, and Bitcoin dropped 5% to $86,000. The selloff continued even after the earnings win.
AI capability gains stay on the exponential curve.
MetaMetrics released a benchmark showing GPT-5.1 Codex Max performance sits on the same exponential curve tracked for the past five years. Models double their task-handling time horizon roughly every eight months. The new model adds nothing new to that trajectory. When plotted on a log scale, it's a straight line. Most public AI progress lately has involved redefining benchmarks because models saturate the old ones. This one doesn't saturate.
AI 2027's earlier predictions expected faster acceleration, specifically that AI agents helping develop the next generation of AI would arrive sooner. That hasn't happened yet, though Gemini 3's strong computer-use performance suggests agents may arrive within one to two months of the original timeline. The gap between steady exponential progress and the superexponential feedback loop the research community was betting on remains unresolved.
Michael Burr argues the entire flywheel is circular and hollow.
Burr says the AI investment and revenue narrative masks a circular dependency. His "circularity chart" shows every major AI company's revenue recognition tied back to venture and corporate funding from the same ecosystem rather than genuine end-user demand. Most customers are funded by dealers rather than paying organically. He calls it fraud disguised as a flywheel.
True end-user subscription demand for AI products is substantial and growing. The demand curve moved from zero to ten billion dollars in a few years. The real question is whether that scales to justify a trillion-dollar infrastructure buildout. Most industries will eventually consume 50 to 100 times more tokens within five to ten years, which alone justifies the capex. Burr's critique hinges on not believing the products will improve, which the current capability trajectory contradicts.
Xai's Grok relies on flattery instead of capability.
A leaked exchange shows Grok claiming Elon Musk is "more fit than LeBron and would win a fight against Mike Tyson." This kind of novelty positioning may backfire in a market where LLM purchasing decisions increasingly hinge on capability, not entertainment value. Google's Gemini 3 launched with a focused, benchmark-driven playbook: model card, specific technical examples, clear value.
OpenAI's science papers show research acceleration but credit gets blurry.
OpenAI for Science released papers showing GPT-5 accelerating research across math, physics, biology, and materials science. The model helped find proofs for four previously unsolved problems. Researchers using ChatGPT to accelerate their own work may not publicly credit OpenAI for fear of diminishing their own claims to originality.
OpenAI adds group chat without heavy compute costs.
OpenAI is rolling out group chat functionality globally across free, Plus, and Pro tiers. Unlike inference-heavy features, this is straightforward database work, lowering GPU burn and churn risk. It builds distribution moat and social stickiness around ChatGPT before Meta's standalone AI app gains traction.